We study two first-order economic consequences of vertical mismatch, using a simple (neoclassical) model of under- and over-employment. Individuals of high type can
perform both skilled and unskilled jobs, but only a fraction of low-type workers can
perform skilled jobs. People have different costs over these jobs, akin to a Roy model.
First, we calibrate the model to match US time-series since the 1970s and show that
changes in educational mismatch has contributed half as much as skilled-bias technological progress for the rise in the college premium. Second, we calibrate the model to match moments of 50 US states, to measure the output costs of frictions generating mismatch. The cost of frictions is 2.5 per cent of output on average but varies between 1.5% to 4% across states. The key variable that explains the output cost of vertical mismatch is not the percentage of mismatched workers but their wage relative to well-matched workers.